]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/downloader/f4m.py
Initiate new release.
[youtubedl] / youtube_dl / downloader / f4m.py
index fc9642905f3fffb5f1ec11f959c36fd486b757f2..c8fde9a89093393132262f7b7d5ec60d83de4b8d 100644 (file)
@@ -3,7 +3,6 @@ from __future__ import division, unicode_literals
 import base64
 import io
 import itertools
-import os
 import time
 
 from .fragment import FragmentFD
@@ -12,37 +11,47 @@ from ..compat import (
     compat_urlparse,
     compat_urllib_error,
     compat_urllib_parse_urlparse,
+    compat_struct_pack,
+    compat_struct_unpack,
 )
 from ..utils import (
-    encodeFilename,
     fix_xml_ampersands,
-    sanitize_open,
-    struct_pack,
-    struct_unpack,
     xpath_text,
 )
 
 
+class DataTruncatedError(Exception):
+    pass
+
+
 class FlvReader(io.BytesIO):
     """
     Reader for Flv files
     The file format is documented in https://www.adobe.com/devnet/f4v.html
     """
 
+    def read_bytes(self, n):
+        data = self.read(n)
+        if len(data) < n:
+            raise DataTruncatedError(
+                'FlvReader error: need %d bytes while only %d bytes got' % (
+                    n, len(data)))
+        return data
+
     # Utility functions for reading numbers and strings
     def read_unsigned_long_long(self):
-        return struct_unpack('!Q', self.read(8))[0]
+        return compat_struct_unpack('!Q', self.read_bytes(8))[0]
 
     def read_unsigned_int(self):
-        return struct_unpack('!I', self.read(4))[0]
+        return compat_struct_unpack('!I', self.read_bytes(4))[0]
 
     def read_unsigned_char(self):
-        return struct_unpack('!B', self.read(1))[0]
+        return compat_struct_unpack('!B', self.read_bytes(1))[0]
 
     def read_string(self):
         res = b''
         while True:
-            char = self.read(1)
+            char = self.read_bytes(1)
             if char == b'\x00':
                 break
             res += char
@@ -53,18 +62,18 @@ class FlvReader(io.BytesIO):
         Read a box and return the info as a tuple: (box_size, box_type, box_data)
         """
         real_size = size = self.read_unsigned_int()
-        box_type = self.read(4)
+        box_type = self.read_bytes(4)
         header_end = 8
         if size == 1:
             real_size = self.read_unsigned_long_long()
             header_end = 16
-        return real_size, box_type, self.read(real_size - header_end)
+        return real_size, box_type, self.read_bytes(real_size - header_end)
 
     def read_asrt(self):
         # version
         self.read_unsigned_char()
         # flags
-        self.read(3)
+        self.read_bytes(3)
         quality_entry_count = self.read_unsigned_char()
         # QualityEntryCount
         for i in range(quality_entry_count):
@@ -85,7 +94,7 @@ class FlvReader(io.BytesIO):
         # version
         self.read_unsigned_char()
         # flags
-        self.read(3)
+        self.read_bytes(3)
         # time scale
         self.read_unsigned_int()
 
@@ -119,7 +128,7 @@ class FlvReader(io.BytesIO):
         # version
         self.read_unsigned_char()
         # flags
-        self.read(3)
+        self.read_bytes(3)
 
         self.read_unsigned_int()  # BootstrapinfoVersion
         # Profile,Live,Update,Reserved
@@ -184,6 +193,11 @@ def build_fragments_list(boot_info):
     first_frag_number = fragment_run_entry_table[0]['first']
     fragments_counter = itertools.count(first_frag_number)
     for segment, fragments_count in segment_run_table['segment_run']:
+        # In some live HDS streams (for example Rai), `fragments_count` is
+        # abnormal and causing out-of-memory errors. It's OK to change the
+        # number of fragments for live streams as they are updated periodically
+        if fragments_count == 4294967295 and boot_info['live']:
+            fragments_count = 2
         for _ in range(fragments_count):
             res.append((segment, next(fragments_counter)))
 
@@ -194,11 +208,11 @@ def build_fragments_list(boot_info):
 
 
 def write_unsigned_int(stream, val):
-    stream.write(struct_pack('!I', val))
+    stream.write(compat_struct_pack('!I', val))
 
 
 def write_unsigned_int_24(stream, val):
-    stream.write(struct_pack('!I', val)[1:])
+    stream.write(compat_struct_pack('!I', val)[1:])
 
 
 def write_flv_header(stream):
@@ -223,6 +237,12 @@ def write_metadata_tag(stream, metadata):
         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 
 
+def remove_encrypted_media(media):
+    return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
+                                 'drmAdditionalHeaderSetId' not in e.attrib,
+                       media))
+
+
 def _add_ns(prop):
     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 
@@ -244,9 +264,7 @@ class F4mFD(FragmentFD):
             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
             if 'id' not in e.attrib:
                 self.report_error('Missing ID in f4m DRM')
-        media = list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
-                                      'drmAdditionalHeaderSetId' not in e.attrib,
-                            media))
+        media = remove_encrypted_media(media)
         if not media:
             self.report_error('Unsupported DRM')
         return media
@@ -293,7 +311,8 @@ class F4mFD(FragmentFD):
         man_url = info_dict['url']
         requested_bitrate = info_dict.get('tbr')
         self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
-        urlh = self.ydl.urlopen(man_url)
+
+        urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
         man_url = urlh.geturl()
         # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
         # (see https://github.com/rg3/youtube-dl/issues/6215#issuecomment-121704244
@@ -303,7 +322,7 @@ class F4mFD(FragmentFD):
         doc = compat_etree_fromstring(manifest)
         formats = [(int(f.attrib.get('bitrate', -1)), f)
                    for f in self._get_unencrypted_media(doc)]
-        if requested_bitrate is None:
+        if requested_bitrate is None or len(formats) == 1:
             # get the best format
             formats = sorted(formats, key=lambda f: f[0])
             rate, media = formats[-1]
@@ -313,7 +332,11 @@ class F4mFD(FragmentFD):
 
         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
-        boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, base_url)
+        # From Adobe F4M 3.0 spec:
+        # The <baseURL> element SHALL be the base URL for all relative
+        # (HTTP-based) URLs in the manifest. If <baseURL> is not present, said
+        # URLs should be relative to the location of the containing document.
+        boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, man_url)
         live = boot_info['live']
         metadata_node = media.find(_add_ns('metadata'))
         if metadata_node is not None:
@@ -340,17 +363,21 @@ class F4mFD(FragmentFD):
 
         dest_stream = ctx['dest_stream']
 
-        write_flv_header(dest_stream)
-        if not live:
-            write_metadata_tag(dest_stream, metadata)
+        if ctx['complete_frags_downloaded_bytes'] == 0:
+            write_flv_header(dest_stream)
+            if not live:
+                write_metadata_tag(dest_stream, metadata)
 
         base_url_parsed = compat_urllib_parse_urlparse(base_url)
 
         self._start_frag_download(ctx)
 
-        frags_filenames = []
+        frag_index = 0
         while fragments_list:
             seg_i, frag_i = fragments_list.pop(0)
+            frag_index += 1
+            if frag_index <= ctx['fragment_index']:
+                continue
             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
             query = []
             if base_url_parsed.query:
@@ -360,24 +387,26 @@ class F4mFD(FragmentFD):
             if info_dict.get('extra_param_to_segment_url'):
                 query.append(info_dict['extra_param_to_segment_url'])
             url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
-            frag_filename = '%s-%s' % (ctx['tmpfilename'], name)
             try:
-                success = ctx['dl'].download(frag_filename, {'url': url_parsed.geturl()})
+                success, down_data = self._download_fragment(ctx, url_parsed.geturl(), info_dict)
                 if not success:
                     return False
-                (down, frag_sanitized) = sanitize_open(frag_filename, 'rb')
-                down_data = down.read()
-                down.close()
                 reader = FlvReader(down_data)
                 while True:
-                    _, box_type, box_data = reader.read_box_info()
+                    try:
+                        _, box_type, box_data = reader.read_box_info()
+                    except DataTruncatedError:
+                        if test:
+                            # In tests, segments may be truncated, and thus
+                            # FlvReader may not be able to parse the whole
+                            # chunk. If so, write the segment as is
+                            # See https://github.com/rg3/youtube-dl/issues/9214
+                            dest_stream.write(down_data)
+                            break
+                        raise
                     if box_type == b'mdat':
-                        dest_stream.write(box_data)
+                        self._append_fragment(ctx, box_data)
                         break
-                if live:
-                    os.remove(encodeFilename(frag_sanitized))
-                else:
-                    frags_filenames.append(frag_sanitized)
             except (compat_urllib_error.HTTPError, ) as err:
                 if live and (err.code == 404 or err.code == 410):
                     # We didn't keep up with the live window. Continue
@@ -397,7 +426,4 @@ class F4mFD(FragmentFD):
 
         self._finish_frag_download(ctx)
 
-        for frag_file in frags_filenames:
-            os.remove(encodeFilename(frag_file))
-
         return True